Video summary
Al bear says we should let OpenAl and Anthropic die
Main summary
Key takeaways
Summary
Ed Zitron argues that Anthropic and OpenAI have economics and funding needs that are difficult to reconcile with their reported revenue. He says their spending commitments make them systemically important to cloud providers and the broader market, but argues that they should not be rescued if they fail. The hosts push back that markets are already repricing risk and that the companies may still obtain financing, albeit at a higher cost.
Key financial claims and figures
- Anthropic: Zitron cites $413 billion in uncanceled liabilities and questions whether the company can meet its commitments from cash flows. He refers to an $80 billion net loss, of which $72 billion was described as “converted”; the transcript is unclear about the accounting treatment. He argues investors need the company’s cash-flow details and GAAP figures, rather than annualized growth rates and adjusted metrics.
- Zitron questions whether Anthropic’s reported computing costs are too low relative to its commitments. He cites a $3.75 billion SpaceX computing deal and about $6 billion in spending on Amazon, Google, and Microsoft computing. The transcript does not make the timing or precise scope of these figures clear.
- He also cites about $9 billion in Anthropic computing costs in a quarter and mentions large additional commitments, including $161 billion associated with Broadcom and a $35 billion deal with Apollo Global Management. The transcript’s details on these agreements are unclear.
- OpenAI: Zitron says the company may need $750 billion by 2030 and refers to a $30 billion funding round. He also claims it spent $12 billion on computing in the first quarter, contrasting that with Anthropic’s reported costs. These are presented as his claims or readings of reports, not independently established figures in the discussion.
- Zitron says Microsoft’s annual operating expenses are about $150 billion and cites a $1.5 trillion funding ask to Middle Eastern investors, who he says face their own capital constraints.
- Market exposure: Zitron claims OpenAI and Anthropic account for 70–80% of total computing demand and more than 70% of revenue for Microsoft, Google, and Amazon. He also cites consensus estimates that cloud services will generate 70% of Microsoft’s growth in coming years, and UBS analyst Stephen Ju’s estimate that OpenAI and Anthropic could account for 48% of Google Cloud revenue.
- CoreWeave: The discussion cites rising bond yields, including 8.3% on some of its lowest-rated bonds and 7.75% on its highest-rated bonds. Zitron says 75–80% of CoreWeave revenue depends on OpenAI and Anthropic. The hosts counter that higher borrowing costs do not necessarily mean financing is unavailable.
- Nvidia: Zitron says Nvidia is targeting $670 billion in fiscal 2028, beginning in February 2027, and argues that the market’s growth expectations ultimately depend on continued GPU sales at rising volumes and prices.
- Generative-AI advertising: eMarketer is cited as forecasting about $6 billion in global revenue by 2030.
Framework and investment considerations
What Zitron would examine in Anthropic’s S-1
- Search first for cash flow and trace how money moves through the company.
- Review how adjusted financial figures are classified.
- Focus on GAAP metrics and balance-sheet clarity, not annualized growth rates or non-GAAP adjustments.
- Compare reported cash costs with the scale and timing of computing commitments.
Risks highlighted
The discussion raises concerns about uncertain cash flows, large and potentially non-cancelable commitments, reliance on external financing, unmonetized users, and data-center delays. The speakers also discuss GPUs sitting unused while data centers are unfinished, potentially postponing payments but not eliminating the underlying obligations.
Credit and market risk
Zitron says OpenAI and Anthropic could merit CCC-level credit ratings, though neither is described as having a rating at the time of the discussion. He argues that funding at such a rating could be difficult and that borrowing $10–20 billion annually would exceed what their cash flows could support. The hosts counter that weaker borrowers may still raise debt, but at a higher price.
Zitron also argues that even a rescue would not necessarily prevent a downturn if hyperscalers’ growth and GPU purchases slowed. He opposes government support and says the companies should be allowed to fail; the hosts raise the potential market-wide consequences of that position.
Recommendations and cautions
Zitron calls for public disclosure of cash flows and understandable GAAP numbers, and cautions against relying on annualized growth rates or adjusted figures. No explicit “not financial advice” disclaimer appears in the subtitles.
Presenters and sources mentioned
Presenters and guest: Ed Zitron (Founder and CEO, EZPR); Julie Hyman; Jake Conley.
Sources cited in the discussion: The Wall Street Journal, The Information, UBS analyst Stephen Ju, Semaphore, eMarketer, and Bloomberg’s Eric Balchunas.
Rate this summary
Your feedback will help improve summaries.
Improve this summary
Reprocess with a stronger model when the summary feels incomplete or inaccurate.
Translate summary in another language
Ask questions to this video
Chat for follow-up questions, clarifications, and source-backed answers.